Calibration method, device and equipment for safe driving state of driver and medium

Through the processing of real-time infrared image information, the characteristic parameters of the head and eye mouth are calculated, and the comprehensive calibration of the driver's safe driving status is achieved, which solves the problem of incomplete calibration in the existing technology, and achieves high efficiency and accuracy driver status evaluation effect.

CN119027920APending Publication Date: 2024-11-26WUHAN JIMU INTELLIGENT TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410988684.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The existing driver's safe driving status calibration method cannot perform comprehensive and efficient calibration. It only relies on the speed threshold, head attitude angle and line of sight direction, and cannot comprehensively evaluate the driver's safe driving status, resulting in the inability to effectively judge fatigue driving and division. Heart driving behavior.

Method used

By obtaining the driver's real-time infrared image information, face detection and face key point detection, calculating the head 3D posture angle, eye and mouth feature point information, determining the status characteristic parameters of the eyes and mouth, combining the mean value Smooth filtering processing and weight updates realize all-round calibration of the driver's safe driving status.

Benefits of technology

It realizes all-round, efficient and accurate calibration of the driver's safe driving status, and can judge the driver's fatigued driving and distracted driving behavior in real time and adaptively, without considering changes in installation and driving sitting posture, which is more in line with the actual driving situation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027920A_ABST
    Figure CN119027920A_ABST
Patent Text Reader

Abstract

The invention provides a driver safe driving state calibration method and device, equipment and a medium, and relates to the technical field of safe driving, and the method comprises the steps: obtaining the real-time state infrared image information of a driver, and carrying out the face detection and face key point detection, calculating head 3D attitude angle information, eye key point information and mouth region feature point information according to a detection result; according to the set features required for judging the dangerous driving behavior, the eye key point information and the mouth region feature point information, determining each state feature parameter, needing to be calibrated, of the eyes and the mouth; the change amplitude of the head 3D attitude angle information and the eye opening degree in a set time period are counted, and a calibratable state is determined; according to the calibratable state, calibrating each state characteristic parameter, needing to be calibrated, of the eyes and the mouth, and judging the dangerous driving behavior of the driver according to each state characteristic parameter calibration value. According to the invention, all-directional accurate calibration of each state under the safe driving condition of the driver is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of safe driving and calibration detection technology, and in particular to a method, device, equipment and medium for calibrating a driver's safe driving status. Background Art

[0002] With the continuous development of autonomous driving technology and advanced driver assistance systems (ADAS), ensuring the safety of drivers during driving has become increasingly important. In order to effectively monitor and evaluate the driver's status and avoid potential risks such as fatigue driving and distracted driving, the calibration of the driver's safe driving status has become an indispensable link. According to the driver's safe driving status, the driver's state characteristic parameters under normal driving conditions can be obtained, based on which the driver's dangerous behavior can be more accurately judged.

[0003] At present, researchers have not paid much attention to the calibration of driving status. The only calibration method is to calibrate the driver's head posture angle and line of sight direction when the vehicle speed meets the threshold. In the existing calibration method, the precondition of calibration is whether the vehicle speed meets the threshold, and there is no calibration for the full speed range; at the same time, the calibration content only involves the head posture angle and line of sight direction. The calibration content is too little to calibrate the driver's status in a safe driving situation.

[0004] Therefore, a comprehensive and efficient calibration method is needed to calibrate the driver's state under safe driving conditions, so as to accurately judge the driver's fatigue driving and distracted driving behavior. Summary of the invention

[0005] In view of this, an embodiment of the present application provides a method for calibrating a driver's safe driving status, so as to achieve the purpose of comprehensively, efficiently and accurately calibrating the driver's safe driving status.

[0006] The embodiment of the present application provides the following technical solution: a method for calibrating a driver's safe driving state, comprising:

[0007] Acquire the real-time infrared image information of the driver, perform face detection and face key point detection on the infrared image information, and obtain face key point information;

[0008] Based on the facial key point information, calculate and obtain the head 3D posture angle information, eye key point information and mouth area feature point information;

[0009] According to the set features required for judging dangerous driving behavior, as well as the eye key point information and the mouth area feature point information, determine the state feature parameters that need to be calibrated for the eyes and the mouth;

[0010] Count the change amplitude of the head 3D posture angle information and the eye opening degree O within the set time period t0. If the change amplitude of the head 3D posture angle information is less than the set amplitude value, and the average value of the eye opening degree O is greater than the set threshold, it is determined that the calibration condition is met, and the current state is determined as a calibrable state;

[0011] According to the determined calibrable states, the characteristic parameters of each state of the eyes and mouth that need to be calibrated are calibrated, and the driver's dangerous driving behavior is judged according to the calibrated values ​​of each state characteristic parameter.

[0012] According to an embodiment of the present application, face detection and face key point detection are performed on the infrared image information to obtain face key point information, including:

[0013] The driver's face position information is obtained through the RetinaFace network, and the face boundary is expanded to a set range according to the face position and size to obtain the face image area;

[0014] The face image area is used as an input image for face key point detection, input into a face key point detection network model, and the face key point information is output.

[0015] According to an embodiment of the present application, the face key point detection network model adopts a distillation training method, wherein the teacher network model in the distillation training method adopts the HRnet-w32 network, and the student network model adopts the MobilenetV3-large network.

[0016] According to an embodiment of the present application, based on the set features required for judging dangerous driving behavior, as well as the eye key point information and the mouth area feature point information, the state feature parameters of the eyes and mouth that need to be calibrated are determined, including:

[0017] According to the eye key point information, the distance T between the upper and lower eyelids, the eye opening degree O, and the curvature values ​​C0 and C1 of the upper and lower eyelids are calculated; according to the mouth area feature point information, the distance ratio R between the left and right mouth corners and the mouth center point is calculated.

[0018] According to an embodiment of the present application, the calibration method further includes: before calibrating each state feature parameter that needs to be calibrated for the eyes and mouth, performing mean smoothing filtering on the state feature parameters to remove abnormal values ​​in the process of solving each state feature parameter itself, wherein the formula of mean smoothing filtering is as follows:

[0019] P t =(P t-1 +P t+1 ) / 2 (a)

[0020]

[0021] The mean smoothing filter processing is performed on the characteristic parameter values ​​of each state in turn through formulas (a) and (b), where t is the time, P t-2 , P t-1 , P t , P t+1 , P t+2 and P t+3 are the time series values ​​of the characteristic parameters of each state.

[0022] According to an embodiment of the present application, each state characteristic parameter that needs to be calibrated for the eyes and mouth is calibrated according to the determined calibrable state, and the calibration method is: within a set time period t0, the mean of the each state characteristic parameter is solved to obtain the calibration value of each state characteristic parameter.

[0023] According to an embodiment of the present application, the calibration method further includes: after calibrating each state feature parameter that needs to be calibrated for the eyes and mouth, while the calibration conditions are continuously met, updating the calibration values ​​of each state feature parameter with set weights.

[0024] The present application also provides a device for calibrating a driver's safe driving status, comprising:

[0025] An image acquisition and processing module is used to obtain the driver's real-time infrared image information, perform face detection and face key point detection on the infrared image information, and obtain face key point information;

[0026] A feature information calculation module is used to calculate and obtain head 3D posture angle information, eye key point information and mouth area feature point information based on the face key point information;

[0027] A calibration parameter calculation module, used to determine the characteristic parameters of each state of the eyes and mouth that need to be calibrated according to the set features required for judging dangerous driving behavior, as well as the eye key point information and the mouth area characteristic point information;

[0028] A calibratable state determination module is used to count the change amplitude of the head 3D posture angle information and the eye opening degree O within a set time period t0. If the change amplitude of the head 3D posture angle information is less than the set amplitude value, and the average value of the eye opening degree O is greater than the set threshold, it is determined that the calibration condition is met, and the current state is determined as a calibratable state;

[0029] The state parameter calibration module is used to calibrate the various state characteristic parameters that need to be calibrated for the eyes and mouth according to the determined calibrable states, and to judge the driver's dangerous driving behavior according to the calibration values ​​of the various state characteristic parameters.

[0030] The present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for calibrating the driver's safe driving status when executing the computer program.

[0031] The present application also provides a computer-readable storage medium, which stores a computer program for executing the above-mentioned method for calibrating the driver's safe driving status.

[0032] Compared with the prior art, the beneficial effects that can be achieved by at least one of the above technical solutions adopted in the embodiments of this specification include at least the following: Compared with the current calibration method, the embodiments of the present invention realize the calibration of more parameters, and at the same time, the precondition of the calibration is to use the state of the driver when driving safely as a reference, which is more in line with the actual situation, and realizes the full-scale calibration of various states of the driver when driving safely. The method of the embodiment of the present invention can perform real-time and adaptive calibration during driving, without considering the changes in installation and driving posture, and achieves the effect of full-scale and efficient calibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 It is a schematic diagram of a method for calibrating a driver's safe driving state according to the present invention;

[0035] Figure 2 It is a schematic flow chart of an embodiment of a method for calibrating a driver's safe driving state of the present invention;

[0036] Figure 3 is a schematic diagram of a device for calibrating a driver's safe driving state according to an embodiment of the present invention;

[0037] Figure 4 It is a structural schematic diagram of the computer device of the present invention. DETAILED DESCRIPTION

[0038] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0039] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0040] like Figure 1 As shown, an embodiment of the present invention provides a method for calibrating a driver's safe driving state, comprising:

[0041] S101. Obtaining the real-time infrared image information of the driver, performing face detection and face key point detection on the infrared image information, and obtaining face key point information;

[0042] S102. Calculate and obtain head 3D posture angle information, eye key point information, and mouth area feature point information based on the facial key point information;

[0043] S103. Determine the characteristic parameters of each state of the eyes and mouth that need to be calibrated according to the set features required for judging dangerous driving behavior, as well as the key point information of the eyes and the characteristic point information of the mouth area;

[0044] S104. Count the change amplitude of the head 3D posture angle information and the eye opening degree O within the set time period t0. If the change amplitude of the head 3D posture angle information is less than the set amplitude value, and the average value of the eye opening degree O is greater than the set threshold, it is determined that the calibration condition is met, and the current state is determined as a calibrable state;

[0045] S105. According to the determined calibrable states, calibrate the characteristic parameters of each state of the eyes and mouth that need to be calibrated, and judge the driver's dangerous driving behavior according to the calibrated values ​​of each state characteristic parameter.

[0046] When S101 is implemented, the real-time status information of the driver, including a single-channel infrared image, is first obtained through a monitoring device such as a host and an infrared camera lens, and the facial key point information is obtained through face detection and facial key point detection algorithms. Specifically, it includes: obtaining the driver's facial position information through the RetinaFace network, and expanding the boundary of the face to a set range according to the position and size of the face to obtain a facial image area; using the facial image area as an input image for facial key point detection, inputting it into a facial key point detection network model, and outputting the facial key point information.

[0047] In one embodiment, the facial key point detection network model adopts a distillation training method, wherein the teacher network model in the distillation training method adopts the HRnet-w32 network, and the student network model adopts the MobilenetV3-large network, and finally outputs 72 key points of the face.

[0048] The RetinaFace network is an advanced face detection algorithm that achieves accurate face positioning and feature point detection by combining multi-task learning and self-supervised learning. RetinaFace uses a multi-task learning strategy to simultaneously predict the face score, face frame, five facial feature points, and the 3D position and correspondence of each facial pixel; and adopts a self-supervised learning algorithm. The algorithm introduces a self-supervised grid decoder branch for parallel prediction of pixel-level 3D shape facial information, which improves the accuracy of face detection.

[0049] The face key point detection network model adopts a distillation training method. The distillation training method (Knowledge Distillation, referred to as KD) is a model compression and acceleration technology that aims to transfer the knowledge of a large and complex model (Teacher Model) to a small and fast model (Student Model). This method can help the student model learn outputs similar to the teacher model, even if the number of parameters of the student model is small; and it can make the student model as close to the performance of the teacher model as possible while maintaining a small model size and computational efficiency. The teacher network model in this embodiment adopts the HRnet-w32 network, and the student network model adopts the MobilenetV3-large network. The HRnet-w32 network is a variant based on the deep learning network structure and belongs to the High-Resolution Network (HRNet) series. This network design is mainly used for various visual tasks, including human posture estimation, image classification, target detection and facial tagging. The MobilenetV3 network is an efficient deep neural network model optimized for mobile devices and is suitable for tasks such as image classification, target detection, and semantic segmentation. The MobilenetV3 network combines hardware-aware network architecture search (NAS) and NetAdapt algorithms and architectural design to reduce latency and computation while maintaining accuracy. MobilenetV3-large is optimized for high-resource usage scenarios. MobilenetV3-large improves the accuracy of ImageNet classification tasks by 3.2% over MobileNetV2, while reducing latency by 20%.

[0050] During the specific implementation of S102, the 3D posture angle information of the head, the eye key point information and the feature point information of the mouth area are calculated based on the facial key point information. Among them, the 3D posture angle information of the head is pitch (pitch angle, the angle of rotation around the Y axis), yaw (yaw angle, the angle of rotation around the Z axis) and roll (roll angle, the angle of rotation around the X axis). The specific calculation process includes: according to the facial key point information, such as facial feature points such as eyes, nose, mouth, ears, etc., a 3D facial model is established based on these facial feature points, the detected 2D facial key points are aligned with the feature points of the 3D facial model, the posture estimation algorithm is used to calculate the 3D rotation of the head, and then the rotation matrix is ​​converted into pitch (pitch angle), yaw (yaw angle) and roll (roll angle).

[0051] like Figure 2As shown, in S103, based on the set features required for judging dangerous driving behavior, as well as the eye key point information and the mouth area feature point information, the various state feature parameters that need to be calibrated for the eyes and mouth are determined. In one embodiment, the method includes: calculating the distance T between the upper and lower eyelids, the eye opening degree O, and the curvature values ​​C0 and C1 of the upper and lower eyelids based on the eye key point information; and calculating the distance ratio R between the left and right mouth corners and the center point of the mouth based on the mouth area feature point information.

[0052] In this embodiment, image processing and computer vision libraries, such as OpenCV, are used to perform key point detection, contour tracking and measurement. Specifically, based on the detected eye key point information, a curve fitting technique is used to track the contours of the upper and lower eyelids, and the distance T between the upper and lower eyelid contour points on the same horizontal line is calculated. The degree of opening and closing of the eyes O is obtained by measuring the vertical distance between the upper and lower eyelids when the eyes are closed, wherein the measurement is performed on the horizontal line of the pupil center or the eye center during the specific measurement. The curvature values ​​C0 and C1 of the upper and lower eyelids of the eyes are obtained by fitting the eyelid contour points to a circle or ellipse and calculating the radius of curvature of the circle or ellipse.

[0053] In the specific implementation of S104, when the driver is in a safe driving state, the direction of his sight is the direction of vehicle travel and he remains focused. The characteristics at this time are that the head posture angle remains stable and the eye opening degree remains large. Based on this, the prerequisite for calibration can be set as follows: within a period of time t0, the maximum change amplitude of the head posture angles pitch, yaw and roll are all less than the set amplitude value Δ, and the average value of the eye opening degree O is greater than the set threshold value α. When this condition is met, it is in a calibrable state.

[0054] In one embodiment, the calibration method further includes: before calibrating each state feature parameter that needs to be calibrated for the eyes and mouth, performing mean smoothing filtering on the state feature parameters to remove abnormal values ​​in the process of solving each state feature parameter itself, wherein the formula of mean smoothing filtering is as follows:

[0055] P t =(P t-1 +P t+1 ) / 2 (a)

[0056]

[0057] The mean smoothing filter processing is performed on the characteristic parameter values ​​of each state in turn through the above formulas (a) and (b), where t is the time, P t-2 , P t-1 , P t , P t+1 , P t+2 and Pt+3 are the time series values ​​of the characteristic parameters of each state.

[0058] The data after smoothing and filtering can be calibrated. In one embodiment, according to the determined calibrable state, the state characteristic parameters that need to be calibrated for the eyes and mouth are calibrated. The calibration method is: within the set time period t0, the mean of the state characteristic parameters is solved to obtain the calibration value of each state characteristic parameter.

[0059] In one embodiment, the calibration method further includes: after calibrating each state feature parameter that needs to be calibrated for the eyes and the mouth, while the calibration conditions are continuously met, updating the calibration values ​​of each state feature parameter with set weights.

[0060] In the specific implementation, the parameters after successful calibration are updated with a weight w=0.1 when the calibration conditions are continuously met, specifically:

[0061]

[0062] Among them, P t is the state value of each parameter at the current time t, and They are the calibration values ​​of each parameter at time t-1 and time t respectively. So far, the calibration of each state parameter has been completed, based on which the driver's dangerous driving behavior is judged.

[0063] The method for calibrating the driver's safe driving state in the embodiment of the present invention obtains the driver's real-time state information through a monitoring device, obtains the head posture angle information and eye key point information through face detection and face key point detection algorithms, and counts the change amplitude of the head pitch angle, yaw angle and roll angle within a period of time. When the change amplitude of the three is very small, it can be determined that this is a calibrable state under the safe driving situation, and calibrates the state values ​​of the driver's head posture angle, upper and lower eyelid distance, curvature and opening degree, and the ratio of the distance between the left and right mouth corners and the center of the mouth that characterizes the facial deflection degree, and judges whether the driver has dangerous behaviors such as fatigue driving and distracted driving according to the calibration values ​​of each state parameter. The calibration values ​​of each parameter are updated with a certain weight in the calibrable state.

[0064] Compared with the current calibration method, this method can calibrate more parameters. At the same time, the precondition of calibration is to use the driver's safe driving state as a reference, which is more in line with the actual situation. In addition, this method can perform real-time and adaptive calibration during driving, without considering the changes in installation and driving posture, achieving a comprehensive and efficient calibration effect.

[0065] like Figure 3As shown, the present application also provides a calibration device 200 for a driver's safe driving state, comprising:

[0066] The image acquisition and processing module 201 is used to obtain the real-time infrared image information of the driver, perform face detection and face key point detection on the infrared image information, and obtain face key point information;

[0067] The feature information calculation module 202 is used to calculate the head 3D posture angle information, eye key point information and mouth area feature point information according to the face key point information;

[0068] The calibration parameter calculation module 203 is used to determine the characteristic parameters of each state of the eye and mouth that need to be calibrated according to the set characteristics required for judging dangerous driving behavior, as well as the eye key point information and the mouth area characteristic point information;

[0069] The calibratable state determination module 204 is used to count the change amplitude of the head 3D posture angle information and the eye opening degree O within a set time period t0. If the change amplitude of the head 3D posture angle information is less than the set amplitude value, and the average value of the eye opening degree O is greater than the set threshold, it is determined that the calibration condition is met, and the current state is determined as a calibratable state;

[0070] The state parameter calibration module 205 is used to calibrate each state characteristic parameter that needs to be calibrated for the eyes and mouth according to the determined calibrable state, and judge the driver's dangerous driving behavior according to the calibration value of each state characteristic parameter.

[0071] During specific implementation, the image acquisition and processing module 201 obtains the driver's facial position information through the RetinaFace network, and expands the boundary of the face by a set range according to the position and size of the face to obtain a facial image area; the facial image area is used as an input image for facial key point detection, input into a facial key point detection network model, and the facial key point information is output.

[0072] In a preferred embodiment, the face key point detection network model adopts a distillation training method, wherein the teacher network model in the distillation training method adopts the HRnet-w32 network, and the student network model adopts the MobilenetV3-large network.

[0073] In specific implementation, the calibration parameter calculation module 203 calculates the distance T between the upper and lower eyelids, the eye opening degree O, and the curvature values ​​C0 and C1 of the upper and lower eyelids according to the eye key point information; and calculates the distance ratio R between the left and right mouth corners and the center point of the mouth according to the mouth area feature point information.

[0074] In one embodiment, the calibration device 200 of this embodiment further includes a data processing module, which is used to perform mean smoothing filtering on each state characteristic parameter that needs to be calibrated for the eyes and mouth before calibrating the state characteristic parameters, and eliminate abnormal values ​​in the process of solving each state characteristic parameter itself, wherein the formula of mean smoothing filtering is as follows:

[0075] P t =(P t-1 +P t+1 ) / 2 (a)

[0076]

[0077] The mean smoothing filter processing is performed on the characteristic parameter values ​​of each state in turn through formulas (a) and (b), where t is the time, P t-2 , P t-1 , P t , P t+1 , P t+2 and P t+3 are the time series values ​​of the characteristic parameters of each state.

[0078] In specific implementation, the state parameter calibration module 205 calibrates the state characteristic parameters that need to be calibrated for the eyes and mouth according to the determined calibrable state. The calibration method is: within the set time period t0, the mean of the state characteristic parameters is solved to obtain the calibration value of each state characteristic parameter.

[0079] In one embodiment, the calibration device 200 of this embodiment also includes a calibration value updating module, which is used to update the calibration values ​​of each state feature parameter that needs to be calibrated for the eyes and mouth with set weights when the calibration conditions are continuously met.

[0080] In one embodiment, a computer device is provided, such as Figure 4 As shown, it includes a memory 301, a processor 302 and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned method for calibrating the driver's safe driving status is implemented.

[0081] Specifically, the computer device may be a computer terminal, a server or a similar computing device.

[0082] In this embodiment, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program for executing the above-mentioned method for calibrating the driver's safe driving status.

[0083] Specifically, computer-readable storage media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer-readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable storage media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0084] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of the present invention are not limited to any specific combination of hardware and software.

[0085] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for calibrating a driver's safe driving status, characterized in that: include: Acquire the real-time infrared image information of the driver, perform face detection and face key point detection on the infrared image information, and obtain face key point information; Based on the facial key point information, calculate and obtain the head 3D posture angle information, eye key point information and mouth area feature point information; According to the set features required for judging dangerous driving behavior, as well as the eye key point information and the mouth area feature point information, determine the state feature parameters that need to be calibrated for the eyes and the mouth; Count the change amplitude of the head 3D posture angle information and the eye opening degree O within the set time period t0. If the change amplitude of the head 3D posture angle information is less than the set amplitude value, and the average value of the eye opening degree O is greater than the set threshold, it is determined that the calibration condition is met, and the current state is determined as a calibrable state; According to the determined calibrable states, the characteristic parameters of each state of the eyes and mouth that need to be calibrated are calibrated, and the driver's dangerous driving behavior is judged according to the calibrated values ​​of each state characteristic parameter.

2. The method for calibrating the driver's safe driving status according to claim 1, characterized in that: Performing face detection and face key point detection on the infrared image information to obtain face key point information includes: The driver's face position information is obtained through the RetinaFace network, and the face boundary is expanded to a set range according to the face position and size to obtain the face image area; The face image area is used as an input image for face key point detection, input into a face key point detection network model, and the face key point information is output.

3. The method for calibrating the driver's safe driving status according to claim 2, characterized in that: The face key point detection network model adopts a distillation training method, wherein the teacher network model in the distillation training method adopts the HRnet-w32 network, and the student network model adopts the MobilenetV3-large network.

4. The method for calibrating the driver's safe driving status according to claim 1, characterized in that: According to the set features required for judging dangerous driving behavior, as well as the eye key point information and mouth area feature point information, the state feature parameters of the eyes and mouth that need to be calibrated are determined, including: According to the eye key point information, the distance T between the upper and lower eyelids, the eye opening degree O, and the curvature values ​​C0 and C1 of the upper and lower eyelids are calculated; according to the mouth area feature point information, the distance ratio R between the left and right mouth corners and the mouth center point is calculated.

5. The method for calibrating the driver's safe driving status according to claim 1, characterized in that: The calibration method further includes: before calibrating each state characteristic parameter that needs to be calibrated for the eyes and the mouth, performing mean smoothing filtering on the each state characteristic parameter to eliminate abnormal values ​​in the process of solving each state characteristic parameter itself, wherein the formula of mean smoothing filtering is as follows: P t =(P t-1 +P t+1 ) / 2 (a) The mean smoothing filter processing is performed on the characteristic parameter values ​​of each state in turn through formulas (a) and (b), where t is the time, P t-2 , P t-1 , P t , P t+1 , P t+2 and P t+3 are the time series values ​​of the characteristic parameters of each state.

6. The method for calibrating the driver's safe driving status according to claim 1, characterized in that: According to the determined calibrable state, the state characteristic parameters that need to be calibrated for the eyes and mouth are calibrated. The calibration method is: within the set time period t0, the mean value of the state characteristic parameters is solved to obtain the calibration value of the state characteristic parameters.

7. The method for calibrating the driver's safe driving status according to claim 1, characterized in that: The calibration method further includes: after calibrating each state feature parameter that needs to be calibrated for the eyes and the mouth, when the calibration conditions are continuously met, updating the calibration values ​​of each state feature parameter with set weights.

8. A device for calibrating a driver's safe driving status, characterized in that: include: An image acquisition and processing module is used to obtain the driver's real-time infrared image information, perform face detection and face key point detection on the infrared image information, and obtain face key point information; A feature information calculation module is used to calculate and obtain head 3D posture angle information, eye key point information and mouth area feature point information based on the face key point information; A calibration parameter calculation module, used to determine the characteristic parameters of each state of the eyes and mouth that need to be calibrated according to the set features required for judging dangerous driving behavior, as well as the eye key point information and the mouth area characteristic point information; A calibratable state determination module is used to count the change amplitude of the head 3D posture angle information and the eye opening degree O within a set time period t0. If the change amplitude of the head 3D posture angle information is less than the set amplitude value, and the average value of the eye opening degree O is greater than the set threshold, it is determined that the calibration condition is met, and the current state is determined as a calibratable state; The state parameter calibration module is used to calibrate the various state characteristic parameters that need to be calibrated for the eyes and mouth according to the determined calibrable states, and to judge the driver's dangerous driving behavior according to the calibration values ​​of the various state characteristic parameters.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for calibrating the driver's safe driving status according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the method for calibrating the driver's safe driving state according to any one of claims 1 to 7.